On Mon, Dec 2, 2013 at 12:03 PM, Lee Jones [off-list ref] wrote:

quoted

quoted

+Optional properties:
+  - st,syscfg          : Phandle to boot-device system configuration registers
+  - st,boot-device-reg : Address of the aforementioned boot-device register(s)
+  - st,boot-device-spi : Expected boot-device value if booted via this device
+
+Example:
+       spifsm: spifsm at fe902000{
+               compatible         = "st,spi-fsm";
+               reg                =  <0xfe902000 0x1000>;
+               reg-names          = "spi-fsm";
+               pinctrl-0          = <&pinctrl_fsm>;
+               st,syscfg          = <&syscfg_rear>;
+               st,boot-device-reg = <0x958>;
+               st,boot-device-spi = <0x1a>;

I don't think we should encode any register offsets whatsoever in device
tree but maybe that's just me. (Yes, Stephen will beat me up about
pin control single, but I consider that a special case.)

I would just put the last two things as #defines into the driver file(s)
or - if it's related to other syscfg registers and varies with SoC incarnation,
as a #define in a shared header for that syscfg thing.

No can do. This isn't _this_ device's register offset, this is a
syscfg register offset which a) there is no driver to apply specific
register offsets to and b) are liable to change oversubsequent SoCs.

So it can be in <linux/mfd/my-sysconfig-regs.h> as
a

#define MY_SYSCON_V1_BOOT_DEV_REG 0x958
#define MY_SYSCON_V1_BOOT_DEV_SPI 0x1a

The kernel should know which SoC is in use and
act apropriately right?

But indeed, what you're doing has been done before
in drivers/regulator/anatop-regulator.c where some
register offset is read in from the attribute
"anatop-delay-reg-offset". So it may be that the DT people
need to overrule me on this one :-)

I don't think we should hard-code knowledge of specific
register locations into the device tree at all as that makes
things very hard to debug.

Yours,
Linus Walleij